This sure as Mr Bojangles still grieved seems like SARS-2 is doing the same.
Some corrective factors:
A) How many have the virus, but did not go to a hospital, did not get tested at all.
B) It takes time to die, so the ones that died got the Virus over 2 weeks ago. In this case the death rate should be the number of cases from 2 weeks ago, corrected upward by Factor A.
C) The initial kill rate may be higher, because at the beginning the weakest people are both getting infected, especially through old folks homes and the like, and they are also succumbing to the Virus quicker.
I haven't visited the Italy Data in 3 days, so my model was predicting three days ahead....my model was running consistently hot, but only by a few percent. Italy is winning faster than the model. So here is the presentation
The third column is the "Exponent" The model was reducing it by .995 per day, I need to ramp that up to a bigger drop. I tuned model to actual and a multiplier of .9885 brings the model to within .37% of actual. See far bottom for new Model Predictions.
Of course this "correcting on the fly" will make my model seem better than it originally was. So be it....the goal is to have a series of multipliers that can be applied to the USA data and make better predictions.